- The University of California, Berkeley, saw a 48% increase in applications from first-year computer science students.
- Despite advances in generative AI, students remain keen to pursue careers in computer science.
- Building something new requires human developers.
One of the most persistent concerns about generative AI is that the technology will cause workers to lose their jobs, a notion that has been particularly prevalent in the field of software coding.
Github Copilot can write a ton of code these days, but is it even worth learning computer science now? This is a question that has been on the minds of math-curious high school students since ChatGPT came out in 2022.
There's a new data point that helps answer at least part of this question: Students are still lining up in droves to study computer science at college.
Eye-opening data points
Take the example of the University of California, Berkeley, which is one of the top or near-top universities in computer science.
First-year applications to UC Berkeley’s Department of Computing, Data Science, and Sociology (CDSS) increased by 48% this year. For the fall 2024 entering class, there were 14,302 applicants (non-transfer students) to these CDSS majors. This is an increase from 9,649 applicants the previous year.
By the way, the number of freshman applications to the University of California, Berkeley overall was not much different from last year.
This was announced last week by Professor Jennifer Chayes, Dean of Berkeley's CDSS College, who presented these startling statistics during a fireside chat with Governor Gavin Newsom and Stanford University Professor Fei-Fei Li at the Joint California Summit on Generative AI in San Francisco.
There is a role for human software developers
We then contacted John Denero, a professor of computer science at the University of California, Berkeley, to discuss the matter further.
He's also the chief scientist at generative AI startup Lilt, and previously worked as a Google researcher on Google Translate, one of the first successful AI-powered consumer apps.
“Students are concerned that generative AI will impact the software engineering job market, especially at the entry level, but are still interested in careers in computing,” he told Business Insider in an email. “I tell them that I don't think many of the more challenging aspects of software development can be performed reliably by generative AI today, and that human software developers will continue to play a central role for the foreseeable future.”
AI is not good at new things
Generative AI is now very good at replicating parts of software programs that have been written many times before, DeNero explained.
This includes computer science homework – see this BI article for more information on the extent to which ChatGPT is being used for homework cheating.
What if you want to build something new? Again, you need smart human coders. (This makes sense, since AI models are trained on data; models often run into problems when that information doesn't already exist or is not part of the training dataset.)
Generative AI “requires a lot of thoughtful human intervention to produce something new, and every resulting software development project involves a significant amount of novelty,” DeNero said. “That's the hard part of computing, and that's the interesting part, and it currently requires smart, well-trained people.”
“Generative AI can speed up the more mundane parts of software development, and software developers tend to be quick to adopt efficiency tools,” he added.
What happens in Lilt?
This is also true of what's happening at Lilt, which is building an AI platform for translators.
Google Translate first appeared 18 years ago, and still today, human linguists have jobs and are relied upon when translation really matters. For example, you might be able to use Google Translate to read a train timetable in Japan, but would you use it to translate your most important business contract without having a human expert check it? Probably not.
“While expert human linguists are still at the heart of the process to ensure publication-quality translations, Lilt's task-specific generative AI models make their work much faster, more accurate, and more consistent,” DeNero says. “The result is more texts being translated into more languages with high quality.”
He predicts the same pattern will hold true in software development: that small teams of highly trained human developers will be better able to build useful, high-quality software.
“So future Berkeley graduates will have many opportunities to use their computing skills to improve the world,” DeNero said. “We hope that many of them will end up working at Lilt.”
